MANAGERIAL AND DECISION ECONOMICS Manage. Decis. Econ.

20: 241 257 (1999)

The Substitutability of Brands

Gordon R. Foxall*
Departments of Management and Psychology, Keele Uni6ersity, Staffordshire, UK A behavior analytical interpretation of consumer brand choice is proposed. The brands studied are of fast-moving consumer goods (FMCG) which are functionally substitutable but differentiated by marketing activity (branding). If functional equivalence was the sole basis of brand substitutability, consumers would either practice single brand purchasing or allocate purchases approximately indifferently (and approximately equally) among competing brands. Brand shares would tend at equilibrium toward equal size. But consumers typically buy several brands of those available, while brand shares vary markedly. The observed patterns of consumer choice are consistent with the predictions of matching theory if brands are assumed to be imperfect substitutes that provide qualitatively different reinforcers. Copyright 1999 John Wiley & Sons, Ltd.

INTRODUCTION Ehrenberg (1988, for example) has shown that comparatively few purchasers of a product category are 100%-loyal to a particular brand. Most consumers show multi-brand purchasing over a sequence of shopping opportunities, choosing within a small repertoire of available brands. Ehrenberg explains this in terms of the functional similarities of brands within a product category. Usually, they have near-identical physical formulations and perform identical tasks. The consumer typically exchanges one brand for another because the benefits gained from one are directly substitutable with those provided by others within the repertoire. Ehrenberg is equally known for his approach to theory formulation, resolutely opposing premature theory-building, and favoring instead the detailed observation of buyer behavior and the empirical determination of patterns within the data, especially recurrences of brand choice regularities. Theory can, and should, come later (Ehrenberg, 1993). However, Ehrenbergs assumptions about, and explanations of, the patterns he identifies can be evaluated only if there is a theoretical account of consumer choice that elucidates the meaning of his findings. His work, and that of his colleagues, collaborators and other re* Correspondence to: Departments of Management and Psychology, Keele University, Staffordshire ST5 5BG, UK.

searchers, spans several decades: it is surely time to move on now to the theory-building phase of this research program. In this paper, I draw attention to the resemblance of sequential patterns of brand choice to matching, a phenomenon that has been extensively analyzed by behavior analysts (notably and originally by Herrnstein, 1961, 1970, 1997; for an extensive review, see Davison and McCarthy, 1988a). The aim is to unite marketing research and behavior analysis by accounting for consumers choice of brands as a function of the pattern of rewards they confer. According to behavior analysis, behavior is predicted and controlled by environmental events rather than intentional processes. The rate at which a response is repeated is a function of the rewarding (or, more accurately, reinforcing) and punishing consequences it has previously generated. The fundamental explanatory tool is the three-term contingency in which a cue (or discriminative stimulus) sets the occasion for a response, which produces rewarding and/or punishing consequences. These consequences, rather than internal deliberation and decision-making, explain the behavior. Behavior analysis has been successfully applied to the study of economic behaviors in both human and nonhuman animals and is an important component of a school of theoretical and empirical research in behavioral economics and economic psychology.

CCC 01436570/99/050241-17$17.50 Copyright 1999 John Wiley & Sons, Ltd.

242

G.R. FOXALL

Behavior analysis and behavioral economics can make important contributions to our understanding of patterns of brand choice. Brand purchasing is a behavior that involves choice among a number of alternative forms of the product category. Matching is a relationship between a pattern of behavioral choices among alternatives and the pattern of reward to which those choices lead. In view of the lack of a theoretical account of why multi-brand purchasing takes the form it does, the paper proposes that matching and me lioration (the theoretical behavioral process believed to underlie matching) provide a framework within which consumer choice may be better understood.1 The paper also demonstrates the difficulties of using matching, which is delineable with some precision in experimental settings, in behavioral interpretations. It proposes a definition for matching that is appropriate for consumer research and shows that matching theory gives rise to two interpretations of consumer choice, which require further experimental analysis. Equally, marketing analysis can contribute to behavior analysis by pointing to the marketingbased factors that maintain repeat purchasing, a set of contingencies more complex than behavior analysts usually deal with. The task for the behavioral interpretation of human consumption is to provide an account of patterns of distributed choice for brands of fast-moving consumer goods (FMCG), which are functionally substitutable but differentiated by marketing activity (branding). Behavioral economics has proceeded at the level of the product category (concerning itself with goods or commodities) and has considered functional performance to be the sole source of reinforcement. It thus overlooks the role of brand differentiation in real-world consumer markets. The current discussion on the nature of, and the motivation for, patterns of consumer choice suggests that the concept of reinforcement, a central component of behavior analysis, requires further analysis.

of responding becomes the dependent variable. Herrnsteins (1961) initial discovery was that when animals are presented with two opportunities to respond (pecking key A or key B), each of which delivers either reward or reinforcement (access to a food hopper) on its own variable interval (VI) schedule,2 they allocate their responses to A and B in proportion to the rates of reward available in A and B. In other words, response rate (B1) is proportional to the relative rate of reinforcement (R ) (de Villiers and Herrnstein, 1976): Bx /(Bx + By ) = Rx /(Rx + Ry ). (1)

This phenomenon, which Herrnstein (1961, 1970) refers to as matching, provides a framework for the behavioral analysis of consumption. As long as there are no differences among reinforcers in terms of bias, i.e. preference for one reinforcer based on characteristics such as its physical position or color, and sensiti6ity, i.e. responsiveness to the alternative reinforcers, Equation (1) simplifies to Bx /By = Rx /Ry. (2)

Taking bias and sensitivity into account, Baum (1974) proposed the generalized matching law: Bx /By = b (Rx /Ry )s (3)

where B is the behavior allocated to alternatives x and y, R is the reinforcers contingent upon that behavior, and the constants b and s represent bias and sensitivity, respectively. Bias and Sensitivity Bias is absent when b, which is the intercept when Equation (3) is re-expressed in logarithmic form, equals unity. Deviations of b from unity indicate a consistent preference for one alternative over the other(s) regardless of the reinforcement rates in operation. As long as the reinforcements for each of the available responses are apparently equal and predict a behavioral indifference between them, a measure of b greater or less than one indicates that preference is biased by some unknown, but invariant, asymmetry between the alternatives (Baum, 1974, p. 2333). Bias is the result of a deficiency in experimental design rather than a shortcoming of the experimental subject; it represents a failure to take account of all of the independent variables that influence preference and declines as relevant independent variables are Manage. Decis. Econ. 20: 241257 (1999)

CHOICE AS BEHAVIOR Matching Herrnstein (1997) defines choice not as an internal deliberative process but as a rate of intersubjectively observable events that are temporally distributed. In his analysis, the relative frequency Copyright 1999 John Wiley & Sons, Ltd.

THE SUBSTITUTABILITY OF BRANDS

243

increasingly taken into account (Baum, 1974). Principal sources of bias are undetected response costs imposed in the case of one alternative but not the other(s), such as an additional effort required to shift one lever in an experiment, and a qualitative difference between reinforcers, such as an unanticipated additional value accorded to one reinforcer but not to the rest (Baum, 1979; Pierce and Epling, 1983; Davison and McCarthy, 1988b). The behavior of a subject who disproportionately chooses the leaner schedule of reinforcement (i.e. who chooses it more often than strict matching would predict) is said to exhibit undermatching; in such cases the exponent s, slope, is less than one. The behavior of a subject who disproportionately chooses the richer schedule of reinforcement exhibits overmatching, and s \ 1. Low sensitivity to reinforcement schedules may arise because the subject is unable to discriminate between the alternatives sufficiently well, especially if there is no delay in reinforcement when responses are allocated to a new choice (and are, therefore, controlled by a different schedule), and because rates of deprivation differ between the schedules (Baum, 1974, 1979). The generalized matching law can thus take a variety of data into consideration (Green and Freed, 1993). In contrast to the stimulus-response psychologies, Herrnsteins matching equation represents response frequencies as a function of reinforcement frequencies. The resulting choice rule indicates that the average reinforcement rate of response A comes to equal the average reinforcement rate of response B. Melioration The empirical findings on matching have been extrapolated from the laboratory to provide an interpretation of complex human economic behavior in terms of melioration, the process in which a difference between local rates of reinforcement leads to a continuous change in the distribution of behavior in the direction of an equality of local reinforcer rates (Davison and McCarthy, 1988a, p. 136). An everyday example of melioration involves the way in which drivers on a major highway frequently switch lanes, selecting the clearest and fastest way forward, returning to the original lane or a third when that becomes the most advantageous. Overall, the driver may or may not reach the final destination more quickly than had he/she remained in the one lane for the Copyright 1999 John Wiley & Sons, Ltd.

entire journey, but immediate advantage (the local rate of reinforcement) leads to an averaging of the rates of reinforcement over all choices. An equilibrium is finally reached when the average reinforcement rates of each lane are equalized. Where T1 and T2 are the times allocated to the two responses, the local difference in reinforcer rates, Rd, is: Rd = (R1/T1) (R2/T2). (4)

Time allocation changes as a result of the sensitivity of behavior to local rates of reinforcement; stabilization is achieved when Rd = 0. Melioration, in which the behavior offering the immediately higher or highest rate of reinforcement is chosen, may result in particular circumstances in the overall maximization of reinforcement, but usually leads to a sub-optimal outcome. Melioration thus provides a molecular level mechanism to explain the behaviors to which matching refers (Herrnstein and Vaughan, 1980).

MARKETING, SUBSTITUTABILITY AND MATCHING Brands and Branding A goal of behavior analysis is the interpretation of complex human behaviorsthose not amenable to direct experimental analysis according to principles derived from studies of responding in simpler contexts (e.g. Skinner, 1969, 1988). As Rachlin (1976, p. 569) argued, even when matching is not directly observed it may still be convenient to assume that matching underlies all choice. Hence, matching becomes a deductive, rather than inductive lawnot a principle derived from experimentation, but a premise with which experimental findings can be interpreted (p. 570). This working hypothesis does not ignore the multiple sources of influence on consumer behavior. Hursh (1980) wrote nearly 20 years ago that Because reinforcers differ in elasticity and because reinforcers can be complementary, no simple unidimensional choice rule such as matching can account for all choice behavior (pp. 219220). Matching is not the only behavioral framework for dealing with the complexity of choice, as the growth of behavioral economics demonstrates (e.g. Green and Kagel, 1987). Marketing analysis, moreover, has identified a number Manage. Decis. Econ. 20: 241257 (1999)

244

G.R. FOXALL

of influences on consumer choice, which may complicate matching theory in this context. Consumer behavior is shaped by a host of marketing influencessuch as branding, advertising, sales promotions, and distribution strategies and by social pressures not under the direct control of marketers, which determine the plasticity of demand (Penrose, 1959) and its responsiveness to changes in these non-price variables. Matching the tendency of individual organisms to allocate responses among alternatives in proportion to the reinforcement obtained from eachis a well-documented phenomenon of both non-human and human responses in experimental contexts (Davison and McCarthy, 1988a). Matching is a molar process (i.e. concerned with the relationship of the frequency of responding to frequency of reinforcement) which is identifiable from comparison of the rates at which responses are emitted and reinforcement obtained. Herrnstein (1979) pioneered its explanation in terms of the molecular process (i.e. concerned with explaining an individual response), melioration, in which the behavioral option offering the higher local rate of reinforcement is chosen at any time; equilibrium is reached when responses are allocated so as to equalize the average reinforcement rates. The extension of matching to the interpretation of non-experimental consumer behavior in humans is commonplace (e.g. Rachlin and Laibson, 1997). Its further extension to the interpretation of consumers distributed brand choices within a product category appears, on the face of it, to be a straightforward matter. (A product category is the set of functionally equivalent brands, each member of which embodies all of the essential functional characteristics of the category; indeed, it must do so in order to hope to compete with established brands. Branding is the differentiation of brands within a product category by means of managerial action.) Brand choices within a product category, such as the selection of either the Heinz or the Crosse and Blackwell brand from a range of baked beans products on a supermarket shelf, follow well-documented patterns (Ehrenberg, 1988). The impulse to interpret consumers sequential brand purchasing in terms of matching underlain by melioration is, therefore, compelling. Moreover, the clearest evidence for the matching law comes from experiments in which the alternative reinforcers are direct substitutes for Copyright 1999 John Wiley & Sons, Ltd.

one another (Davison and McCarthy, 1988a); Heyman (1996) argued that both perfect substitutability and confidence that the nominal reinforcement frequencies exclusively control behavior are required for matching to occur.3 Green and Freed (1993) argued that substitutability inheres in the similarity of the functional attributes of the reinforcers (goods or commodities). Yet in affluent consumer markets, manufacturers and retailers annually incur large expenditures not only on production systems and quality controls to ensure that the physical formulation of their brand is standard for the product category, but also on branding and promotional efforts to differentiate their brand(s) from those of other manufacturers and retailers. While the former expenditure is fully explicable in terms of Green and Freeds (1993) understanding of the substitutability of reinforcers as consequences of purchase that provide a set of functional benefits, the latter expenditure can be understood in behavior analytic terms only as an extension of the meaning of reinforcement. These considerations illustrate the kinds of assumption and procedure that a behavioral interpretation of consumer choice needs to adopt. Green and Freed (1993, p. 151) point out that work on the matching law generally has used reinforcers that are qualitatively similar (indeed identical) reinforcers. They also note that in choices between qualitatively different reinforcers (such as between orange juice and grapefruit juice), relative obtained reinforcement value would not equal relative amount consumed; yet if one assumes the matching relation to be true, then some other factor must be incorporated to preserve the relation between relative obtained reinforcement value and relative amount consumer for qualitatively different reinforcers (Green and Freed, 1993, 151). Rachlin et al. (1980) go so far as to claim that substitutability inheres in the measure of sensitivity, s, of the generalized matching law (Equation (3)); s = 1 would, therefore, imply perfect substitutability. They adduce empirical evidence for this view, showing that in the case of pigeons, rats and monkeys choices of food versus water, s : 10 indicating complementary products, whereas for these animals choices of food versus food and water versus water, s : 1. Although this is not a universally accepted view, there is general agreement even among its critics that s represents qualitatively Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

245

different reinforcers (Baum and Nevin, 1981). While economists have generally studied nonsubstitutes, psychologists have concentrated on substitutes. The assumption of both has been that highly branded versions of a product category (Coca Cola and Pepsi Cola, for instance) are substitutes. The integration of matching research and behavioral economics is desirable in order to combine their ideas of substitutability. A marketing analysis raises additional questions such as what is the relationship between brands of this sort and those less-differentiated by marketing activity? (such as own-label colas). The Meaning of Matching in Marketing The idea of matching needs to be carefully defined in the context of buyer behavior. It is important to recognize that the matching law says nothing about consumption. All reinforcers obtained are assumed to be consumed. If matching implied simply that the proportion of buying responses for Brand A equals the proportion of reinforcers obtained from that brand, it would be a truism. Assuming that the reinforcement value from consuming a commodity is constant, and that all commodities purchased are consumed (and these seem to be reasonable assumptions), then the proportion of purchases would always match the proportion of reinforcers. The application of matching to marketing must avoid this tautology. The problem derives in part from the fact that matching was developed on interval schedules (where reinforcement rate can be used as an independent variable) and purchasing behavior is based on a ratio-like schedule.4 The true independent variable on a ratio-like schedule (price) does not translate nicely into the independent variable of the matching law (reinforcement rate) because reinforcement rate is dependent on response rate on ratio schedules. As a result, translations between the matching law and consumer behavior are not straightforward. More suitable variables for consumer behavior have the advantage that they are readily measurable, however. In the analysis of consumer behavior, an appropriate unit of choice (i.e. the dependent variable) is spending, not purchasing. Spending would be measured in monetary units such as dollars or pounds. An appropriate unit of reinforcement is the number of actual purchases made, given a particular ratio of spending (i.e. Copyright 1999 John Wiley & Sons, Ltd.

price per unit by volume, weight or size). This is not a true independent variable because it is determined by spending patterns. Unfortunately, this is a byproduct of using the matching law with ratiolike schedules. With these adjustments, the matching law comes to state that The proportion of dollars/ pounds spent for a commodity will match the proportion of reinforcers earned (i.e. purchases made as a result of that spending. Frequency of purchase is thus the independent variable. This avoids the problem of tautology (there are possible conditions under which this would not be true). It also avoids the problem of having varying amounts of reinforcement from each act of consumption. The predicted equilibrium point for behavioral allocation on concurrent variable ratio schedules is exclusive choice of the richest schedule, i.e. that with the lower or lowest ratio requirement (Herrnstein and Loveland, 1974; Herrnstein and Vaughan, 1980), and this has been borne out empirically (Green et al., 1983). The definition captures the essence of a market transaction. We do not know, from the data on multibrand buying, what the precise pricing schedules are. However, by assuming that functional utility is (a) the sole operative reinforcer (the homogene ity assumption), and (b) constant from brand to brand (the constancy assumption), we can ascertain which is the leaner or richer schedules from relative brand/unit prices. The aggregate data of marketing analyses are then invaluable for indicating how often consumers switch brands and how far they allocate responses between the leaner and richer schedules.

Assumptions and Predictions The brands in a product category are substitutes inasmuch as they are functionally interchangeable; usually, they are of near-identical physical formulation. A new brand, in order to become accepted as a member of the product category into which it is introduced, must incorporate the functional attributes offered by existing members (Ehrenberg, 1991). Assuming homogeneity of reinforcement and constancy of reinforcement across brands, because the schedules involved are conc VR VR, the first prediction is that consumers will show exclusi6e purchase of the lowest priced alternati6e. Manage. Decis. Econ. 20: 241257 (1999)

246

G.R. FOXALL

There is an important exception to the prediction and finding that, for consumers faced with conc VR VR schedules, maximizing reinforcement is the only option consistent with matching. Matching, as has been argued, applies to identical or near-identical reinforcers substitutes. If, however, choices are non-substitutable, the prediction is of nonexclusive responding on conc VR VR schedules (Green and Freed, 1993, p. 152). Hence, the second prediction is that, if the brands are not perfect substitutes, consumers will purchase se6eral brands within a product category. In view of the ubiquity of branding in consumer product markets, the third prediction is that the source of any non -exclusi6e purchasing will be found in the branding acti6ities of firms. This prediction, based on an assumption commonly found in the marketing literature, involves a relationship whose nature needs to be made specific because it is the essence of the behavior analytic account of the substitutability of brands.

PATTERNS OF BRAND CHOICE Exclusive Choice: The Sole Purchasers A small proportion of buyers is loyal to one brand over a sequence of 10/15 purchases of the product category (Table 1). Each brand attracts exclusive purchasers who are a relatively small proportion of buyers of the product category. Larger, more differentiated brands attract a rather higher proportion of exclusive purchasers than small brands. Non-exclusive Choice: The Multi-brand Purchasers Most customers of any brand buy other brands far more often than they buy it. Table 2 shows that a coffee customer typically makes three purchases of the brand per year but nine purchases of the product category; each brand displays more or less the same pattern. Similar patterns are found for brands in other product categories: for example, American consumers of breakfast cereals, which tend to be highly-differentiated, make about five purchases of a brand per year, but 35 purchases of the product category; British consumers of gasoline, a product category which is much more difficult to differentiate, make ten purchases of a given brand annually, but 50 of the product category (Ehrenberg and Goodhardt, 1979). From Copyright 1999 John Wiley & Sons, Ltd.

these figures can be calculated the annual share of category requirements (SCR), which is the average number of brand purchases divided by the average number of product category purchases over a year. Breakfast cereals show an SCR of 13%; for gasoline, the SCR is 20%. Whatever the brand, its customers buy a similar range of other brands and do so in a replicated pattern. Thus, Table 3 indicates that Maxwell House was, on average, bought by 41% of the customers of each of the other brands, and Maxim by about 12% of each other brands customers. These phenomena appear predictable from the penetration (and market share) levels of each brand: hence, Maxwell Houses penetration was the highest; Maxims was the lowest. Finally, it is notable that, apart from those relatively few customers who are 100%-loyal to any brand, buyers tend to restrict their purchases to a small subset of brands rather than spreading them across the entire brand set. Even 100% brand loyal buyers are not particularly heavy buyers of their preferred brand. The variability in consumers choices is also borne out by the data on penetration rates and market shares, which diverge markedly from brand to brand (Table 2). (Market share records the percentage of product category sales accounted for by each brand. Penetration measures the percentage of potential buyers of a brand who in fact purchased it in a given time period.) For instant coffee, which is typical of consumer product categories, annual brand penetrations range from 6% to 24%, and market shares from 3% to 19%. Table 1 shows similar results for a wider range of products. Repeat purchase loyalty tends to be similar for brands that have similar market shares: compare the repeat rates for highly-differentiated versus less-differentiated brands in Tables 1 and 2. Smaller brands not only attract fewer buyers but those buyers buy less of the brand (or buy it less frequently), a phenomenon known as doublejeopardy (Ehrenberg et al., 1990). Both the SCRs and the number of sole buyers are lower for smaller brands (Ehrenberg and Uncles, 1999). There is no evidence of rigid market partitioning into clusters of brands that exclusively attract some customers rather than others. However, a buyer of one of the more-differentiated brands is more likely to buy another highly-differentiated brand on a subsequent purchase occasion than a less-differentiated brand (Tables 1 and 3). Manage. Decis. Econ. 20: 241257 (1999)

Copyright 1999 John Wiley & Sons, Ltd. Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

Table 1.
Brands

Annual Performance Measures for the Eight Leading Brands

Brand size Market share % Percentage buying 46 36 26 24 16 15 13 8 23 Loyalty Purchases per buyer 4.6 4.1 3.7 3.3 3.7 3.0 3.2 3.9 3.7 Percentage buying once 36 40 47 50 46 55 55 55 48 100% loyal % rate 4 4 4 4 3 3 3 3 3 13 14 15 14 15 15 15 16 15 Category purchases/buyer Partitioning % who also bought 1st 52 57 56 54 56 57 52 55 3rd 32 33 36 35 36 36 32 34 5th 19 20 22 24 21 20. 20 21

1st 2nd 3rd 4th 5th 6th 7th 8th Average

27 19 12 9 7 5 4 3 8

20 15 10 10 11 9 8 7 11

Source: Ehrenberg and Uncles (1999, p. 13). Product categories: Catsup, cereals, cheese, chilled orange juice, gasoline, household cleaners, laundry detergents, paper towels, take-home beer and toothpaste. Based on panel data from Nielsen, IRI, AGB, GfK, TCI.

247

248

G.R. FOXALL

Table 2.

Annual Penetration and Average Purchase Frequencies (Leading Brands in Order of Market Share)
Market share % Penetration % 67 24 21 22 22 18 13 9 6 20 17 Average purchasesa Brand 3.6 3.1 2.8 2.6 2.7 2.9 2.0 2.6 3.0 2.8 Any 7 9 9 9 8 9 11 9 11 9 9

Instant coffee, USA, Annual

Any instant Maxwell House Sanka Tasters Choice High Point Folgers Nescafe Brim Maxim Other Average brand

100 19 15 14 13 11 8 4 3 13 11

a Per buyer of the brand. Source: Ehrenberg and Uncles (1999, p. 6). Although this table refers to instant coffee, the data base is comprehensive: the product categories investigated by Ehrenberg et al. include: 30 food and beverage items ranging from cookies to take-home beer; 20 personal care products and cleaners from cosmetics to washing-up liquids; industrial and durable goods including gasoline, aviation fuel and motor cars; stores, store chains, and shopping trips; and audience viewership patterns for TV programs and channels. The research summarized here was undertaken, between 1950 and 1995, in the UK, Continental Europe, the USA and Japan).

Table 3.

Duplications of Purchases between Brands

Percentage who also bought 1 36 31 34 51 48 33 52 41 2 32 32 38 30 39 45 38 36 3 29 32 34 35 34 39 51 36 4 32 40 36 40 40 44 39 39 5 38 25 28 31 34 27 34 31 6 26 23 20 22 25 20 17 22 7 13 20 17 18 15 15 25 17 8 13 11 14 10 11 8 16 12

Instant coffee, USA, annual Buyers of 1. 2. 3. 4. 5. 6. 7. 8. Maxwell House Sanka Tasters Choice High Point Folgers Nescafe Brim Maxim

Average duplication

Source : Ehrenberg and Uncles (1999).

Branding and Differentiation Surveys and field experiments have shown three relationships between purchasing and its consequences that (a) establish the consequences as genuine reinforcers for the consumer behaviors of interest, and (b) will prove of interest in the interpretation of multi-brand purchasing as matching. The first of these relationships is among price, sales and brand differentiation. In this relationship, Ehrenberg (1986) describes an experiment to identify the relationship of price-toCopyright 1999 John Wiley & Sons, Ltd.

brand differentiation. Eighteen sales calls were made on consumers who were offered three or four brands or versions of each of three products: cookies, leaf tea and ready-to-eat breakfast cereals. The product categories represented several levels of brand differentiation: physically different product formulations (for cookies: tea biscuit versus shortbread; for breakfast cereals: Corn Flakes versus Rice Krispies); a hea6ily ad6ertised brand and a generic : Kelloggs Corn Flakes versus the identical product in a pack simply labelled Corn Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

249

Flakes; marginally -differing 6ersions of the same ad6ertised brand : United Biscuits Rich Tea cookie appeared in both standard form and in an otherwise-identical pack containing the word NEW; the firms Pennywise brand was similarly presented in two versions; and identical generics differentiated only by price, sold simply as Tea but at three different prices. Bimonthly sales calls were made over an 8-month period at the participants homes; the participants, all medium-toheavy users of each product category, formed a quota sample by age and social class in three UK locations. One hundred and eighty-one consumers participated over the entire period, divided into two matched subsamples; participants were allowed to buy none or one pack of each product on each call. The results indicate that, for near -identical brands, e6en a small price differential will result in the higher -priced brand recei6ing only a small share of purchases; howe6er, if the higher -priced brand is markedly differentiated from the other brand in some way (e.g., in fla6or, packaging, ad6ertising, or some other marketing 6ariable ), it can hold a large share of the market. Brand shares tend to be remarkably stable into the medium term and, following major sales upset caused by promotional activities, the shares tend to resettle at or near to their old levels. This experiment identified a number of components of branding that influence sales: package familiarity, levels of past advertising support, different market shares external to the sales test, and individual respondents previous brand preferences or buying habits. The author concludes that other elements of the marketing mix, including product quality, availability, branding, promotional and merchandizing activities, relaunches and perhaps media advertising, also seem likely to have demonstrable sales effects that could be identified and measured by experimental means. The second study establishes that a promotional price reduction has an effect on brand sales (albeit smaller than might be expected ) but brand shares and repeat purchase rates return to their pre promotional le6els once the promotion has ended. Ehrenberg et al. (1994) investigated the aftereffects of FMCG promotions based on price in the UK, USA, Germany and Japan. Such promotions are known to increase sales markedly for the duration of the offer; the research question was whether there would be an after-effect of a sales Copyright 1999 John Wiley & Sons, Ltd.

promotion in terms of higher repeat purchase rates. (The authors assumed that sales would rise during the promotion and fall subsequently: indeed, they used a sales blip as evidence for a promotional effect. Their actual research question is more subtle than the mere expectation that sales levels owing to the promotion would not be sustained once the deal had come to an end.) Price promotions for packaged grocery products were examined in the four countries, and data were gathered from various kinds of household panel, ranging from scanner output, diaries and garbage diaries; each national sample consisting of between 1000 and 5000 respondents. Sales peaks deviations of 25% or more from the usual steady rate of purchases typical of these marketswere taken as evidence of major promotions (many peaks exceeded a 50% increase in sales but basing the criterion level of sales increase on this figure did not affect the analysis). Such large sales peaks are generally caused only by promotional campaigns and subsequent checking revealed that price-promotions indeed coincide with the large increases in sales identified. Before and after sales levels could be compared for 175 promotioninduced peaks. Small sales differences, both positive and negative, were detected; across the 25 products examined, the average change in sales level was one percentage point. We interpret this as effectively a nil effect: there was little if any general after-effect on sales (Ehrenberg et al., 1994, p. 15). Moreover, repeat buying rates during the 8-week period immediately following each sales peak were unchanged from those recorded for the 8-week period immediately preceding it. Because the sales peaks were clearly the outcome of large numbers of additional consumers purchasing during the promotional periods, why was the increased brand-specific buying observed during the promotions not maintained afterwards? The reason is that some 70% of buyers during the promotional periods were already consumers of the brand, having bought the brand at least once during the preceding 6 months; 80% had done so in the last year and 93% in the last 2.5 years. The underlying pattern of buyer behavior is that some consumers who already include the promoted brand within their portfolio temporarily switch to it while it is offered at a price reduction; however, promoted brands that are not part of their portfolio are not tried just because they are offered at a discount. The additional Manage. Decis. Econ. 20: 241257 (1999)

250

G.R. FOXALL

buyers during the promotion period did not remain disproportionately high-level users of the promoted brand once the deal had come to an end; rather, they settled back to their customary rate of purchasing the brand. Even so, only some 10%20% of the brands long-term customer base react during a promotion by switching purchases; most do not change their buying habits. They buy less frequently than would be required for them to get involved in the short-term promotion. Price promotions reward (some) existing purchasers of the brand, rather than drawing in new buyers. Unfamiliar brands, even when they are highly promoted, tend to draw increased sales only from customers who already feature the brands in their portfolios. Because the usual sales of these brand are lower than for others, the increased purchases induced by the promotion are also relatively low. The third study indicates that token rewards for repeat -loyalty to brand or store do not affect re peat -buying rates. A particular kind of promotion occurs when retailers attempt to influence repeatbuying rates (or loyalty) by rewarding consumers for making current purchases of a particular brand or at a particular store in the expectation that this will encourage continued patronage. The rewards usually consist of tokens (coupons) redeemable on subsequent purchase occasions when they lead to a reduction in the consumers overall bill. Consumers usual pattern of store choices closely resembles that of their brand choices (Kau and Ehrenberg, 1984). For example, American consumers who in the course of a year bought ground coffee at Safeway purchased it (anywhere) on average 11.1 times. Those who purchased at Safeway did so 3.3 times; they also purchased ground coffee at other retail outlets 7.8 times. Similar patterns of many infrequent buyers and low repeat loyalty are apparent for chains other than Safeway (Uncles and Ehrenberg, 1990). Sharp and Sharp (1997) describe and evaluate a store loyalty program in which panel-generated repeat-buying data were compared with predictions obtained from the Dirichlet model.5 The purpose of a loyalty program is to lock consumers in to a specific brand and might, therefore, be expected to generate a longer-term effect than a one-off price promotion. The propensity of consumers to revert to their baseline patterns of brand choices once a sales promotion has ended has been noted (Ehrenberg et al., 1994). The expectation was that the aggregate sample of conCopyright 1999 John Wiley & Sons, Ltd.

sumer would: (i) switch less often to non-program stores (those not included in the promotion); (ii) increase the program stores shares of total product category purchases; (iii) increase repeatbuying rates for the program stores; (iv) increase usage frequency for these stores; (v) show a greater tendency toward exclusive loyalty to these stores; and (vi) increase their switching among program stores while decreasing their tendency to switch to non-program stores (Sharp and Sharp, 1997). The program Fly Buys is the largest of its type in Australia, and covers all purchases made at participating stores or with the program credit card; in order to accumulate points, consumers are required to present a magnetic strip card when they make payment. Observed variables were close to the Dirichlet predictions; for instance, penetration statistics generally fall within two percentage points of the predictions. The data show no general effect of the Fly Buys program on repeat buying patterns. Penetrations of brands sold at Fly Buys, measured as average purchase frequencies, were not consistently higher or lower than prior to the program; customers did not allocate a higher share of their purchases to Fly Buys brands than they had prior to the program; nor did program brands attract a higher than expected number of sole-purchasers (100%-loyal consumers). Deviations were found for two of the six program brands investigated in that they produced a higher repeat rate than predicted (excess loyalty) but they did so for non-program as well as program members. Ehrenberg et al. (1994) concluded that It is difficult to interpret these results as evidence of Fly Buys changing the repeat purchase patterns of these markets (p. 483).

INTERPRETATION The Focus of Interpretation We have noted that the only behavior consistent with matching on conc VR VR (variable ratio; see note 4) schedules is exclusive choice of the richer/richest schedule. Such maximization is the only strategy consistent with matching in these circumstances (Herrnstein and Vaughan, 1980). In fact, three patterns emerge from the tables, only one of which is consistent with this prediction and even then only partially. This first pattern is ex clusi6e choice or sole buying in which a small Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

251

proportion of product category users is 100%-loyal to one brand. Sole purchasers presumably maximize something, but all brands attract some sole buyers. The strategy of repeatedly purchasing a single brand minimizes costs of search. Sole purchasers of cheaper, less-promoted brands also minimize their absolute monetary outlay on the product category and may be insensitive to those reinforcers that derive from branding. Other consumers exclusively purchase heavily-promoted brands. The second pattern is multi -brand buying in which a consumer chooses among a small group (repertoire) of tried and tested brands rather than over the entire brand set. Consumers who practice repertoire buying might be said to show matching because they apparently select among brands according to some principle of melioration perhaps responding locally to small price differentials or non-price deals. But there is no direct evidence of this on the present assumptions. Because the prediction of exclusive purchasing, based on the VR schedules in operation, is not borne out for all consumers, we may at this exploratory stage ask whether matching phenomena other than maximization elucidate the data. If brands provide reinforcers by virtue of their functional substitutability, and do so equally, the consumer who does not, as predicted, settle on one tried and tested brand could be expected to substitute any and all brands in the product category for one another, apparently at random. The brand-indifferent consumer would distribute purchases approximately equally among the brands (and we might refer to them as equitable brand purchasers). Such equitable choice suggests that all brands in a product category would tend toward similar shares of the market. However, the third tendency is a preference for branded goods those heavily-differentiated through marketing and promotional activities; and there are actually large discrepancies in brand shares.6 By far the most common pattern is multi-brand purchasing across a number of brands, with a tendency to favor the highly-differentiated brands. These repertoire buyers are of particular interest. They cannot be said to maximize but we have noted that their switching among all brands, including both the highly-and the less-differentiated, is superficially suggestive of melioration and matching. If the homogeneity and constancy assumptions hold, why would these consumers not maximize by becoming sole purchasers of the least Copyright 1999 John Wiley & Sons, Ltd.

expensive brands, or become equitable brand purchasers? Matching research suggests two avenues of interpretation: bias and sensiti6ity, the b and s of Equation (3), i.e. the generalized matching law. Bias and under- or overmatching can be difficult to infer even from rigorous experiments designed to explore matching (Davison and McCarthy, 1988b), however, and it is worthwhile examining the problem in more detail before attempting an interpretation in these terms. Brand preferences appear to be remarkably resilient to disruptions in the market environment (such as price and sales promotions). Even adverse price differentials do not shift consumers from those preferences as long as the brand is well-differentiated by advertising or other marketing activity. Nor do price deals induce consumers to buy the promoted brand unless they have been buyers of that brand at some time in the past; even then, only relatively few such consumers buy the brand during the promotion. There is no evidence of continuing excess allegiance to a promoted brand once the deal comes to an end. Differential rewards in the form of token loyalty points are similarly unable to dislodge consumer brand preferences. Consumers who have never bought the promoted brands seem particularly resistant to trying it as a result of a price promotion. Why are the consumers who take advantage of price promotions generally those who are already occasional users of the brand who happen to be replenishing their stock of the product category at the moment of purchase? The answer presumably lies in learning history and melioration. Purchase and consumption of the brand in question have presumably been reinforced on several occasions; the brand delivers the functional attributes of the product category; if replenishment is needed, the consumer will make the least expensive choice; the consumer will make this choice for only as long as the price advantage remains. Three further questions are more difficult to answer as long as the assumption is that functional performance is the sole source of reinforcement. First, why does increased value for money attract so few consumers who have previously used the brand? Second, why do non-buyers of a promoted brand not buy it when it provides the required functional rewards more economically? There is no reason to conclude that the learning histories of either group differ significantly from those of consumers who Manage. Decis. Econ. 20: 241257 (1999)

252

G.R. FOXALL

do buy the promoted brand, as long as reinforcement is assumed to be confined to the functional performance of brands. Third, how are we to account for the behavior of many consumers which is not controlled by the contingencies defined by price and functional reinforcement relationships but, apparently, by the elements of branding? Bias It will be recalled that bias is a consistent preference for one choice alternative independent of the reinforcement rate. Brand loyalty is a consistent preference for one choice alternative which is independent of relative price. Consistent choice of a highly-differentiated brand is exactly this. If consumers choose the high status brand over the cheaper generic brand, this is a clear example of bias. All three sources of bias suggested by Baum (1974) may have analogs in consumer research. Response bias would occur when search costs for a particular brand increase as a result of shelf positioning in stores or of the failure of a supermarket to carry a wide spectrum of available brands. In reality, response bias on these grounds is unlikely to seriously affect the analysis of aggregate data collected over a 3-month period. Moreover, patterns of store choice are similar to those of multi-brand purchasing: exclusive patronage is relatively rare (Kau and Ehrenberg, 1984). No one stores marketing practices are, therefore, likely to bias choice consistently during the measurement period. Discrepancy between scheduled and obtained reinforcement is a more serious candidate for bias. By adopting the homogeneity assumption, we have apparently scheduled functional utility as the sole source of reinforcement. If reinforcement is available from some other source, perhaps related to branding, bias will ensue. As Baum (1974, p. 238) argued, choice between qualitatively different reinforcers resembles choice between different amounts of the same reinforcer. Branding may so enhance a particular alternative, perhaps operating as an establishing operation (Michael, 1982) or by presenting rules in the form of augmentals (Zettle and Hayes, 1982), as to render the intended schedules of little value in interpreting consumer behavior. The question of qualitati6ely different rein forcers deser6es comment in it own right. BrandCopyright 1999 John Wiley & Sons, Ltd.

ing points to reinforcing consequences of ownership and consumption (rather than spending or purchasing), over and above those that inhere in the functional properties of the product category. It appears then that brand differentiation supplies discriminative stimuli for reinforcers that are qualitatively different from functional performance. In fact, a brand always signals two sources of reinforcement: functional utility, the intrinsic properties by which the product reduces physical deprivation, and performance feedback, by which it reduces social deprivation. Utility and feedback are qualitatively different sources of reinforcement, albeit ones which act only in combination. A completely unpromoted brand is unlikely to succeed, no matter how well it provides the functional utility that defines the product category. Established brands, therefore, differ in the amount of differentiation they embody. The capacity of highly-differentiated brands to elicit higher spending would be explicable in terms of their providing different reinforcers, because in addition to functional utility they would confer feedback on the level of social and personal performance that would accrue during consumption. They would increase the opportunities for conspicuous consumption, social status, style, reliability, reputation, and so on; in addition, in as much as the products reputation is inextricably bound up with that of the consumer, they would continue to denote his or her skill, expertise and (socially-defined) judgment as a consumer. Purchase and consumption of the less-differentiated brands might actually diminish these. Sensitivity Given the assumptions of homogeneity and constancy, the pattern of multi-brand purchasing identified by marketing analysis suggests undermatching, because the highly-differentiated brands are generally more expensive. This observation does not, of course, reveal why these consumers do not become either sole or equitable purchasers. It also fails to take branding into consideration. Consumers habit of switching between highly- and less-differentiated brands is difficult to attribute to melioration while homogeneity and constancy of reinforcers is presumed. I should like to propose the counterhypothesis that if it is the case that branding provides or signals qualitatively different Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

253

reinforcers from those inherent in the functional performance of the product, then repertoire buyers are showing overmatching (favoring the richer schedules of reinforcement). The view that many consumers practice overmatching by favoring the more differentiated brands initially seems absurd. Overmatching occurs when the ratio of responses to the choice providing the highest reinforcement rate is greater than the ratio predicted by matching. It is difficult to apply overmatching directly to ratio schedules but one possibility is that overmatching would only be present if the dominant brands also happened to be the lowest in price. An overmatching consumer is spending less on the lower cost alternative. However, because purchasing highlydifferentiated brands is actually likely to incur a price premium, this is not the case. Differentiated brands are in fact offered on a leaner schedule than the less-differentiated brands, and those who purchase them disproportionately are presumably undermatching. By contrast, consumers who favored the less-differentiated brands, whose prices would presumably be lower, would be overmatching, disproportionately allocating their spending to those versions of the product that were presented on the richer schedule. The attractiveness of this interpretation lies in its consistency with the experimental analysis of matching and the behavior analytic concepts that underpin it. Reinforcement is homogeneous and the behavior of those consumers who show a tendency to favor the differentiated brands can simply be put down to bias. In that case, the question raised by the data for behavior analysis is as follows: Why should consumers spend more on these relatively expensive brands if they offer no greater functional benefits than cheaper alternatives? It may simply be the case that the elements of branding are discriminative stimuli for the benefits to be obtained in consumption rather than purchase: they are claims that augment the anticipated pleasures of consumption. However, it may be that the elements of branding are rewards in their own right that provide an additional source of reinforcement to that found in the functional characteristics of the product (Foxall, 1998). If this is so, an interpretation in terms of overmatching seems more plausible. In order to establish that overmatching was the case, it would be necessary to show that the favored brands are characterized by richer schedules than Copyright 1999 John Wiley & Sons, Ltd.

other brands. In addition, because there is no evidence that price advantage plays a part in the long-term level of sales for differentiated brandson the contrary, it suggests that price may be a disincentive to loyalty for such brandsthe additional source of reinforcement must be located elsewhere. In other words, if overmatching is the case, the favored brands must be offered on a richer schedule and, because price is not the source of the richness of these brands schedules, purchase of these brands must be reinforced by means additional to functional performance. The price/ promotion experiments support the view that this additional source of reinforcement inheres in the values provided by branding. The qualitative differences between these reinforcers may lie in their differential satiation rates (Baum, 1979).7 As Baum stated,
When the alternatives produce different reinforcers, differential satiation leads to deviation from matching . . . Differing rates of satiation . . . are probably more common than equal ones . . . Increases in the rate of occurrence of one, therefore, will produce increases in responding for the other the opposite of what occurs in choice between alternatives providing the same reinforcer (Baum, 1979, p. 279).

This is closely related to the definition of the complementarily of goods given above. Perfect substitutability and perfect complementarily are more accurately viewed as poles of a continuum rather than as discrete relationships (Kagel et al., 1995). In light of the large shared component of functional substitutability among the brands in a product category, the most that can be claimed is that brands, considered as combinations of functional attributes and symbolic entities, show less than perfect substitutability. If so, the matching law would reflect that spending on brand i as a proportion of all spending on the product category equals reinforcement (functional, symbolic) for brand i as a proportion of all reinforcement (functional, symbolic) from the product category. Utilitarian and Informational Reinforcement The qualitative differences between economic goods can be explained as follows. Brands within a product category comprise discriminative stimuli that refer to two sources of reinforcement: utility and symbol. Brands differ in combining differing amounts of each type of stimulus. They Manage. Decis. Econ. 20: 241257 (1999)

254

G.R. FOXALL

are generally treated by consumers as substitutable because their functional performances are (near-)identical, but brands differentiated by marketing activity are valued more highly over a stream of purchases than those that are lessdifferentiated. Discriminative stimuli that signal utility and symbolism refer to utilitarian and informational reinforcements, respectively (Foxall, 1990). Utilitarian reinforcement derives from the functionality of the purchased item, its economic and technical benefits, and is contingencybased. Informational reinforcement is symbolic rather than substantive; it inheres in feedback on the social and personal effectiveness of consumer behavior, resulting in the conferral of social status and/or self-esteem. It functions by showing the consumer how well he or she is doing according to some symbolic scale of value. It involves verbal behavior and is rule-based (Foxall, 1997). Such a qualitative difference between the utilitarian aspects of the product and the symbolic references it makes is widespread in economic studies of conspicuous consumption, in organizational psychology, in the applied behavior analysis of ecologically-impacting consumption, and in other fields (Foxall, 1996, 1998). However, it is not generally embraced in behavior analysis, possibly because it seems to require a fundamental conceptual shift. The analysis of consumer brand choice in terms of behavior analysis and behavioral economics suggests that it does not require this.

CONCLUSIONS Summary and Implications Behavior analysis provides empirically-based concepts and conceptual insights to elucidate observed patterns of choice in human consumers. Marketing analyses of consumer choice identify two types of buyer behavior: exclusive purchase of a brand and multi-brand purchasing. Over time, sole buyers may become multi-brand purchasers; those observed may simply buy the product category so infrequently as to rule out multi-brand buying during the time period recorded. The sole buyers behavior is consistent with the matching theory prediction for close substitutes presented on conc VR VR schedules in one respect, inconsistent in another. The prediction is that one choice should be exclusively preferred that presented on the most economical schedule. Copyright 1999 John Wiley & Sons, Ltd.

Sole buyers are, by definition, exclusive purchasers of one brand or another, but each brand in the product category has its own set of exclusive buyers. While the theory predicts that one choice will emerge overall as the only brand selected by the entire sample, every brand conforms to this pattern in some degree. Moreover, the larger, more differentiated brands attract a higher proportion of sole buyers. If, as matching theory predicts, these consumers are maximizing, they are presumably maximizing different preferred combinations of functional and non-functional reinforcers. The multi-brand purchasers also show more than one pattern of preference, which require careful interpretation. Although they purchase several brands (a selected repertoire in the case of individuals, the entire brand set aggregately), they also show disproportionate preference for the highly-differentiated brands. The first suggests indifference among the brands; the second qualifies this by identifying a systematic pattern of preference. This may reflect bias preference based on factors other than the reinforcement contingencies. A local store may stock a limited range of brands; even large supermarkets do not carry all versions of the product. Consumers may simply prefer their baked beans in blue packaging rather than in red or green. Any element of branding and merchandizing might effect some degree of bias. This explanation maintains that only functional utility reinforces and that consumers are indifferent to the schedules on which the competing brands are offered. Price differences, relatively small for affluent customers, may actually have little influence on choice. Hence, preference for promoted brands is accounted for as bias. However, the observed patterns of multi-brand purchasing may yet be owing to sensiti6ity to one or other of the schedules. If functional performance is the only source of reinforcement and all brands are equal in this respect, then consumers are opting for the leaner schedule. They are certainly buying the strongly promoted brands disproportionately and may be undermatching (only an experimental analysis can unequivocally determine the latter by calculating s ). If, however, both functional and non-functional reinforcers control these preferences, the latter deriving from firms branding activities, then consumers disproportionate buying of highly-differentiated brands may be evidence of overmatching. The second Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

255

interpretation supports the conclusion that qualitative differences render brands less than perfect substitutes. The extent of the non-substitutability of brands is extensive: the lack of partitioning indicates that consumers differentiate not only between the broad categories of highly- and lessdifferentiated brands, but among the brands within each of these categories. While branding differences may be relatively small within each of the broadly defined categories, they are sufficient to allow consumers to discriminate their preferences among the brands within each. Further Research The analysis suggests that brands within a product category cannot be treated simply as substitutes. The empirical evidence on multibrand purchasing indicates that many such brands lie closer to the substitutability pole of the substitutability independence continuum. To this extent, Ehrenbergs explanation of multi-brand purchasing in terms of the essential similarity of brands is upheld; in the degree to which competing brands are functionally similar, they will provide ready replacements for one another and brand loyalty (understood as repeat purchasing rates) will be less than 100%. The relationship among the choices inherent in multi-brand purchasing requires elucidation, however. Are the choices made by individual consumers attributable simply to indifference among the brands in a product category, or can they be linked, by matching research, to specific differences in the reinforcement provided by the various brands? In particular, is it possible to relate choice to price differences or to variations in measurable non-price rewards? A second line of inquiry is intimated by the cumulative proportions of consumers who are totally loyal to one brand or another (sole purchasers who show exclusive choice). Table 1 shows that many consumers discriminate among competing brands to the extent that they treat one brand as independent of the others within the product category. What is the relationship of such purchasing to price and non-price sources of reward (or, in terms of the conceptual development proposed above, to utilitarian and informational sources of reinforcement)? The aggregate level of analysis on which this paper rests has provided the assignment for beCopyright 1999 John Wiley & Sons, Ltd.

havior analysis and behavioral economics to unravel the underlying patterns of choice that consumers exhibit in their individual behavior. The required level of analysis for further investigation is that of the intensive study of single subjects, which is characteristic of behavioral research. Research in progress which employs this strategy hints that product categories may indeed be allocated along a continuum of substitutabilityindependence (G.R. Foxall (in preparation) Multi-brand purchasing as matching; consumer choice as melioration). Acknowledgements
I am particularly grateful to the referees. In addition, I thank Andrew Ehrenberg for permission to use the tables, participants in an Open University seminar, and the annual conferences of the Experimental Analysis of Behaviour Group, the European Academy of Marketing, and the Association for Behavior Analysis.

NOTES
1. Skinner (1947), the architect of behavior analysis, proposes a similar theory-building process: identification of the basic data, expression of the relationships among them, and the emergence of new concepts. Behavior analysis has generally failed to move on to the third stage (Dinsmoor, 1995), a state of affairs to which this paper may also bring conceptual development, albeit in a specific area. 2. An interval schedule maintains a constant minimum time interval between rewards (or reinforcements). Fixed interval schedules maintain a constant period of time between intervals, while on a variable interval schedule the time varies between one reinforcer and the next (Catania, 1992). The contingencies just described enable behavioral allocation to be controlled and predicted by concurrent variable interval schedules, usually abbreviated to conc VI VI. 3. Two commodities, X and Y, are substitutes if a reduction in the price of X leads to an increase in the quantity demanded of X and a decrease in the quantity demanded of Y. The usual examples are of highly competitive brands in the same product category such as Coca Cola and Pepsi Cola, Cadburys Dairy Milk and Galaxy Milk Chocolate (Kagel et al., 1995). (Complementarity is the converse: a reduction in the price of X leads to an increase in quantity demanded of Y ; commodities are independent if a change in the price of one has no effect on the quantity demanded of the other.) A marketing analysis would emphasize that price is only one aspect of the marketing mix which might influence such changes in quantity demanded. Thus, few if any brands will be perfect substitutes; differentiation is likely to lead consumers to discriminate between them on non-functional grounds.

Manage. Decis. Econ. 20: 241257 (1999)

256

G.R. FOXALL

4. A ratio schedule is one in which a specified number of responses have to be performed before reinforcement becomes available. Fixed ratio schedules keep the number of required responses equal from one reinforcer to the next; variable ratio schedules allow the required number of responses to change from one reinforcer to the next (Catania, 1992). 5. The Dirichlet model (Goodhardt et al., 1984) assumes that the market is stationary, i.e. sales growth of each brand is small and steady in the medium term, and unsegmented, i.e. the brands are independent rather than grouped. The model represents consumer behavior stochastically at the level of the individual household and estimates three parameters of buying for a product: the total number of purchases, consumers rates of product purchase, and their brand choice probabilities. The model may also be used to estimate patterns of store choice. Its inputs are (i) the penetration and average purchase frequency of the product category and the average brand, and (ii) the market shares of relevant brands. Its output descriptions and predictions refer to any brands performance or loyalty measures, such as penetration, purchase frequency, repeat buying and switching, all of which are assumed to reflect its market share. These factors form the baseline levels from which purchase behavior is expected (at least by marketing managers) to vary from baseline during and following promotional interventions. 6. One possibility is that the overpurchased brands are functionally superior. Because it is unlikely, however, that a brand would be given a functional advantage not reflected in its price, the schedule on which it was offered would remain approximately constant. 7. Of the several sources of deviation from s = 1 proposed by Baum (1979), differential satiation is the most relevant to marketing-influenced purchasing. Poor discrimination between alternati6es is unlikely for experienced consumers dealing with qualitatively different reinforcers; changeo6er delay, an experimental procedure in which no reinforcer is available for a short period after a change from one schedule to another, has no obvious analog in consumer behavior; depri6ation of itself is unlikely to influence affluent consumers; and asymmetrical pausing, which refers to a subjects pausing for different times after reinforcement on each schedule, again has no obvious analog.

REFERENCES
W.M. Baum (1974). On two types of deviation from the matching law. Journal of the Experimental Analysis of Beha6ior, 22, 231242. W.M. Baum (1979). Matching, undermatching and overmatching in studies of choice. Journal of the Experimental Analysis of Beha6ior, 32, 269281. W.M. Baum and J.A. Nevin (1981). Maximization theory: some empirical problems. Beha6ioral and Brain Sciences, 4, 389390.

A.C. Catania (1992). Learning, 3rd edn, Englewood Cliffs, NJ: Prentice-Hall. M. Davison and D. McCarthy (eds) (1988a). The Matching Law: A Research Re6iew, Hillsdale, NJ: Erlbaum. M. Davison and D. McCarthy (1988b). The Matching Law: A Research Re6iew, Hillsdale, NJ: Erlbaum. P.A. de Villiers and R.J. Herrnstein (1976). Toward a law of response strength. Psychological Bulletin, 83, 1131 1153. J.A. Dinsmoor (1995). In the beginning . . . In Modern Perspecti6es on B.F. Skinner and Contemporary Be ha6iorism (edited by J.T. Todd and E.K. Morris), Westport, CT: Greenwood Press, pp. 7 24. A.S.C. Ehrenberg (1986). Pricing and brand differentiation. Singapore Marketing Re6iew, 1, 5 15. A.S.C. Ehrenberg (1988). Repeat Buying: Facts, Theory and Applications, 2nd edn, New York: Oxford University Press/London: Griffin. A.S.C. Ehrenberg (1991). New brands and the existing market. Journal of the Market Research Society, 33, 285 299. A.S.C. Ehrenberg (1993). Theory or well-based results: which comes first? In Research Traditions in Market ing (edited by G. Laurent, G. Lilien and B. Pras), Dordrecht: Kluwer Academic Publishing, pp. 79 108. A.S.C. Ehrenberg and G.J. Goodhardt (1979). Essays on Understanding Buyer Beha6ior, New York: Market Research Corporation of America/J. Walter Thompson. A.S.C. Ehrenberg and M.D. Uncles (1999). Under standing Dirichlet -type Markets, London: South Bank Business School. A.S.C. Ehrenberg, G.J. Goodhardt and P. Barwise (1990). Double jeopardy revisited. Journal of Market ing, 54, 82 91. A.S.C. Ehrenberg, K. Hammond and G.J. Goodhardt (1994). The after-effects of price-related consumer promotions. Journal of Ad6ertising Research, July August, 11 21. G.R. Foxall (1990). Consumer Psychology in Be ha6ioural Perspecti6e, London and New York: Routledge. G.R. Foxall (1996). Consumers in Context: The BPM Research Program, London and New York: Thomson. G.R. Foxall (1997). Marketing Psychology: The Paradigm in the Wings, London: Macmillan. G.R. Foxall (1998). Radical behaviorist interpretation: generating and evaluating an account of consumer behavior. The Beha6ior Analyst, 21, 321 354. G.J. Goodhardt, A.S.C. Ehrenberg and C. Chatfield (1984). The Dirichlet: a comprehensive model of buying behaviour. Journal of the Royal Statistical Society, A147, 621 655. L. Green and D.E. Freed (1993). The substitutability of reinforcers. Journal of the Experimental Analysis of Beha6ior, 60, 141 158. L. Green and J.H. Kagel (1987). Ad6ances in Beha6ioral Economics, Vol. 1, Norwood, NJ: Ablex.

Copyright 1999 John Wiley & Sons, Ltd.

Manage. Decis. Econ. 20: 241257 (1999)

THE SUBSTITUTABILITY OF BRANDS

257

L. Green, H. Rachlin and J. Hanson (1983). Matching and maximizing with concurrent ratio-interval schedules. Journal of the Experimental Analysis of Beha6 ior, 40, 217224. R.J. Herrnstein (1961). Relative and absolute strength of response as a function of frequency of reinforcement. Journal of the Experimental Analysis of Beha6 ior, 4, 267272. R.J. Herrnstein (1970). On the law of effect. Journal of the Experimental Analysis of Beha6ior, 13, 243 266. R.J. Herrnstein (1979). Derivatives of matching. Psy chological Re6iew, 86, 486495. R.J. Herrnstein (1997). The Matching Law: Papers in Psychology and Economics (edited by H. Rachlin and D.I. Laibson), New York: Russell Sage Foundation/ Cambridge, MA: Harvard University Press. R.J. Herrnstein and D.H. Loveland (1974). Hunger and constraint in a multiple schedule. Journal of the Ex perimental Analysis of Beha6ior, 24, 107116. R.J. Herrnstein and W. Vaughan (1980). Melioration and behavioral allocation. In Limits to Action: The Allocation of Indi6idual Beha6ior (edited by J.E.R. Staddon), New York: Academic Press, pp. 143 176. G.M. Heyman (1996). Elasticity of demand for alcohol in humans and rats. In Ad6ances in Beha6ioral Eco nomics. Vol 3. Substance Use and Abuse (edited by L. Green and J.H. Kagel), Norwood, NJ: Ablex, pp. 107132. S.R. Hursh (1980). Economic concepts for the analysis of behavior. Journal of the Experimental Analysis of Beha6ior, 34, 219238. J.H. Kagel, R.C. Battalio and L. Green (1995). Eco nomic Choice Theory: An Experimental Analysis of Animal Beha6ior, Cambridge: Cambridge University Press. A.K. Kau and A.S.C. Ehrenberg (1984). Patterns of store choice. Journal of Marketing Research, 21, 399409. J. Michael (1982). Distinguishing between discriminative and motivational functions of stimuli. Journal of

the Experimental Analysis of Beha6ior, 37, 149 155. E.T. Penrose (1959). The Theory of the Growth of the Firm, Oxford: Oxford University Press. W.D. Pierce and W.F. Epling (1983). Choice, matching, and human behavior: a review of the literature. The Beha6ior Analyst, 6, 57 76. H. Rachlin (1976). Beha6ior and Learning, San Francisco, CA: W.H. Freeman and Co. H. Rachlin and D.I. Laibson (1997). Introduction to Herrnstein, R.J., The Matching Law: Papers in Psy chology and Economics, New York: Russell Sage Foundation/Cambridge, MA: Harvard University Press, pp. 1 14. H. Rachlin, L. Green, J.H. Kagel and R.C. Battalio (1980). Substitutability in time allocation. Psycholog ical Re6iew, 87, 355 374. B. Sharp and A. Sharp (1997). Loyalty programs and their impact on repeat-purchase loyalty patterns. In ternational Journal of Research in Marketing, 14, 473 486. B.F. Skinner (1947). Experimental psychology. In Cur rent Trends in Psychology (edited by W. Dennis), Pittsburgh: University of Pittsburgh, pp. 16 49. B.F. Skinner (1969). Contingencies of Reinforcement: A Theoretical Analysis, Englewood Cliffs, NJ: PrenticeHall. B.F. Skinner (1988). Reply to Mackenzie. In The Selec tion of Beha6ior. The Operant Beha6iorism of B.F. Skinner: Comments and Consequences (edited by A.C. Catania and S. Harnad), New York: Cambridge University Press, pp. 113 114. M. Uncles and A.S.C. Ehrenberg (1990). The buying of packaged goods at US retail chains. Journal of Re tailing, 66, 278 294. R.D. Zettle and S.C. Hayes (1982). Rule-governed behavior: a potential framework for cognitive behavioral therapy. In Ad6ances in Cogniti6e Beha6ioral Research and Therapy (edited by P.C. Kendall), New York: Academic Press, pp. 73 117.

Copyright 1999 John Wiley & Sons, Ltd.

Manage. Decis. Econ. 20: 241257 (1999)